JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 102, NO. D3, PAGES 3805-3818, FEBRUARY 20, 1997
Modeling sea-salt aerosols in the atmosphere
1. Model development
S. L. Gong and L. A. Bartie
Atmospheric Environment Service, Downsview, Ontario, Canada
J.-P. Blanchet
Earth Sciences Department, University of Quebec at Montreal, Canada
Abstract. A simulation of the processes of sea-salt aerosol generation, diffusive transport, transformation, and removal as a function of particle size is incorporated into a one-
dimensional
version of the Canadian general climate model (GCMII). This model was
then run in the North Atlantic between Iceland and Ireland during the period of January- March. Model predictions are compared to observations of sea-salt aerosols selected from a review of available studies that were subjected to strict screening criteria to ensure their representativeness. The number and mass size distribution and the wind dependency of total sea-salt aerosol mass concentrations predicted by the model compare well with observations. The modeled dependence of sea-salt aerosol concentration in the surface
layer
(X,/xg m -3) on 10-m
wind speed
(U•0, m s -•) is given
by X = beaU•ø.
Simulations
show that both a and b change with location. The value a and b range from 0.20 and 3.1 for Mace Head, Ireland to 0.26, and 1.4 for Heimaey, Iceland. The dependence of X on surface wind speed is weaker for smaller particles and for particles at higher altitudes. The
residence
time of sea-salt
aerosols
in the first atmospheric
layer (0-166 m) ranges
from 30 min
for large particles (r = 4-8/xm) to -60 hours for small particles (r = 0.13-0.25/xm).
Although some refinements are required for the model, it forms the basis for comparing the simulations with long-term atmospheric sea-salt measurements made at marine
baseline observatories around the world and for a more comprehensive three-dimensional modeling of atmospheric sea-salt aerosols.
1. Introduction
Sea-salt aerosols play a very important role in a variety of
processes in the atmosphere. They influence radiative transfer directly by scattering solar radiation and indirectly by altering
cloud droplet size distribution and concentration, thereby in-
fluencing the albedo of marine boundary layer clouds. In ad- dition, sea-salt aerosol particles are chemical carriers of spe- cies containing C1, Br, I, and S and therefore play a role in the
atmospheric cycles of these important elements. The halogens
Br and C1, once mobilized by heterogeneous reactions from
sea-salt inorganic forms to reactive gaseous forms (e.g., Br 2,
C12) [e.g., Mozurkiewcz, 1995], can play a role in atmospheric
ozone depletion and destruction of light hydrocarbons [Jobson
et al., 1994].
Numerous experimental investigations of sea-salt aerosols
have been conducted around the globe in order to determine
their abundance and physical/chemical properties [Jacobs,
1937; Woodcock, 1953; Toba, 1965a, b; Lovett, 1978; Prospero,
1979; Podzimek, 1980; Parungo et al., 1986; Marks, 1990; Ike- gami et al., 1994; McGovern et al., 1994]. Sea-salt aerosol con- centration is a strong function of the state of the sea surface,
which is in turn determined by meteorological conditions, es-
pecially surface wind speed. Sea-salt particles are distributed
from 0.02 to 60 txm with a bimodal size distribution in the
submicron portion [Fitzgerald, 1991]. Since the observations
Copyright 1997 by the American Geophysical Union. Paper number 96JD02953.
0148-0227/97/96JD-02953509.00
can only be performed at certain times and locations, applica- tion of the experimental results is limited. It is vital to develop an aerosol transport model to predict its concentration and size distribution on a regional scale as a function of atmo-
spheric processes. Various models have been proposed in the
literature [Gathman, 1982; Fairall and Davidson, 1986; Erick- son and Duce, 1988; Fitzgerald, 1992; van Eijk et al., 1992]. A
summary of major functions and assumptions in these models
is shown in Table 1. All of these models are limited to the
marine boundary layer and do not address the long-range
transport of sea salt.
A comprehensive sea-salt aerosol model coupled with a one-
dimensional climate model (FIZ-C) [Therrien, 1993] is pre- sented in this paper. Using the meteorological conditions gen- erated by the FIZ-C, the model includes the following
processes: (1) sea-salt generation due to surface wind; (2)
vertical transport by turbulence and convection; (3) dry depo-
sition and gravitational settling; and (4) wet removal processes
which include both in-cloud and below-cloud scavenging. The
model does not at this time include chemical or physical trans-
formation through interactions with other aerosol types such as
acidic sulphur. The particle size distribution is modeled by representing the size spectrum as a series of discrete size bins.
2. Physical Model
The fate of sea-salt aerosols, once they are injected into the atmosphere from the ocean source, is governed by a series of physical processes such as transport, coagulation, dry and wet
removal, and chemical transformation. Transport of aerosols is
3806 GONG ET AL.: ATMOSPHERIC SEA-SALT MODELING, 1
Table 1. Summary of Sea-Salt Aerosol Models
Size
Authors Dimension Domain Spectrum Transport Source
Removal
Dry Wet
Gathman [1982] ' one MBL no eddy Blanchard [1963]
Fairall and Davidson [1986] one MBL yes eddy Fairall and Davidson [1983]
Burk [1986] one MBL no eddy Monahah et al. [1986]
Stramska [1987] one MSL yes eddy Monahah et al. [1986]
Erickson and Duce [1988] two surface no eddy empirical
FitzgeraM [1992] zero MBL yes eddy Monahah et al. [1986]
van Eijk et al. [1992] zero MBL yes eddy empirical
D • G? D? G? D* G? D? G? D* G* D* G* BC* IC* BC? IC? BC* IC* BC* IC* BC? IC* BC? IC? BC* IC*
MBL, marine boundary layer; D, dry deposition; G, gravitational fall; BC, below-cloud scavenging; IC, in-cloud scavenging.
*Not included.
?Included.
realized by turbulent diffusion, advection, and vertical convec-
tion of the atmosphere, which is represented in this model by
the FIZ-C parameterizations which are analogous to those for
moisture transport in the FIZ-C. A generalized prognostic mass balance equation for size i of type j aerosol particle can
be written as
Opxu
+ divpxuU = S u + I u (1)
Ot
where X'q is the mass mixing ratio of the i th size range aerosols
(kg/kgair),
U is the horizontal
wind
velocity
vector
(m s-•), p is
the air density
(kg m-3), and
Sq is the source
and
sink
terms
which may include following contributions: (1) surface source
(natural and anthropogenic); (2) clear air processes (particle nucleation, particle coagulation, and chemical transforma- tion); (3) in-cloud processes (activation of aerosols, attach-
ment to clouds, and removal of aerosol-attached clouds by
hydrometeors); (4) dry deposition; and (5) precipitation scav-
enging, where particle nucleation, and chemical transforma-
tion are processes not required for sea-salt aerosols. Term I u
represents the rate of intersectional transfer process which
moves the aerosol mass from one size bin to another. The
formation of new aerosol particles through coagulation or breaking of the existing particles is one intersectional transfer
process. Since only dry mass of aerosols is entered into the
mass balance equation (1), the condensation/evaporation of
water on sea-salt aerosol particles will not contribute to the intersectional transfer rate. However, condensation/evapora-
tion may contribute to intersectional transfer rate of other
types of aerosols, for example, sulphate aerosols when sulphu-
ric acid vapor condenses on the existing aerosols.
In accordance with the FIZ-C convention, the prognostic
equation is rewritten in the form of
0-7
-= at
dynamics+-bT
surface clear air+7F
dryOXu OXu
•n-cloud below-clouds
In (2), the aerosol concentration change has been divided into tendencies for dynamics, surface, clear air, dry deposition, in-
cloud and below-cloud processes. The dynamics includes re-
solved motion as well as subgrid turbulent diffusion and con- vection. The surface processes include surface emission rate of
both natural and anthropogenic aerosols and serve as bound-
ary conditions for the model. Particle nucleation, coagulation,
and chemical transformation are included in clear-air process.
In the current version, coagulation is not included for sea-salt
aerosols. However, for purposes of predicting total mass of
sea-salt aerosols for comparison with observations, this is not a major drawback, since Na and C1 which dominate sea-salt aerosol mass are not affected substantially by coagulation. The aerosol model is structured in such a way that any process
which affects the concentration and size distribution of sea-salt
aerosols can be modified or added to the model as new or
better parameterizations for such processes become available.
Some detailed physical parameterizations for sea-salt aerosols
will be presented in the following sections.
2.1. Aerosol Growth With Relative Humidity
Experimental evidence [Fitzgerald, 1991] has shown that there exists a size distribution of sea-salt aerosols of radii ranging from 0.02 to 60 •m. A knowledge of size distribution is necessary in order to calculate the effect of the aerosol on climate as well as on chemical reactions in the atmosphere.
In this model, the aerosol size distribution is simulated by a method similar to that developed by Gelbard and Seinfeld [1980]. The whole size range of aerosol particles is divided into
a number of sections (or size bins), and each size section is
represented by one prognostic equation (i) which is a mass
continuity equation averaged for that size range. The calcula-
tion of sea-salt generation is carried out by integrating the production over each size bin while the physical properties such as dry deposition and wet removal rates are calculated by using the average size in each size bin. It is noted that in the model, the prognostic equation is written for dry size only. Being very hygroscopic, sea-salt aerosols will change size as the
GONG ET AL.' ATMOSPHERIC SEA-SALT MODELING, 1 3807
ticles are always in stable equilibrium with the surrounding relative humidity, the actual size of sea-salt aerosols is related to their dry size by the empirical relationship of Gerber [1985]'
Clr•
2 ] 1/3
C4
__
log
S + r•
(3)
r • C3rd
where r d is the dry particle radius (centimeters), S the satura- tion ratio (or relative humidity), and C1, C2, C3, and C4 are parameters for different types of the aerosol particles (Table 2). Equation (3) can be used in a relative humidity ranging from
0% to 100%.
Physical tendencies, for example, coagulation, dry deposi-
tion, gravitational settling, or wet removal, are computed for each dry size bin using the aerosol size in equilibrium with ambient water vapor. The relative humidity profile is calcu- lated from temperature and moisture predicted by the climate
model.
As the particles grow with relative humidity, the density of
the particles is also altered by incorporating more water. The
average density of sea-salt aerosols in each size bin is calcu- lated based on the size difference between the dry and wet particles. The radius of the particles in equilibrium with am- bient moisture is estimated by (3).
>4 ms"
(b)
wind > 10 rn s"
Figure 1. Mechanisms for sea-salt aerosol generations
(adapted from Monahah et al. [1986]). Two mechanisms are
presented (a) indirect production by bubbles and (b) direct production by spumes.
2.2. Generation of Sea-Salt Aerosols
The generation of sea-salt aerosols has been attributed to
various physical processes [Blanchard, 1983]. The most prom-
inent mechanism is believed to be the entrained air bubbles
bursting during whitecap formations due to the surface wind [Monahah et al., 1986]. Precipitation (both rain and snow)
contributes to the production but only on an intermittent and
local scale. A summary of various mechanisms are shown in
Figure 1.
Physically, the production of sea-salt aerosols by wind is
proportional to the whitecap coverage. Continuous supply of
excessive energy by the wind to the sea surface results in wave breaking. Wu [1986] developed a relationship between the whitecap coverage and wind speed by considering that the energy lost by wave breaking is balanced by the energy gained
from the wind:
E: ß = (CoVo)(C •0 ,•,0) - (4) where W is the fraction of the sea surface with whitecap cov-
erage,
E is the energy
supply
rate by the wind
to a unit area
of
the sea surface, r is the wind stress, V is the sea surface driftcurrent, u, is the wind-friction velocity, U• o is the wind speed at 10-m level, and C lo is the wind stress coefficient which is
proportional to Ul/o 2 [Wu, 1969]. On the basis of shipboard
photographic observations of sea surface [Monahah, 1971;
Table 2. Constants for Growth Equation (3)
Aerosol Model C 1 C 2 C 3 C 4 Sea salt 0.7674 3.079 2.573 x 10-1• - 1.424 Urban 0.3926 3.101 4.190 x 10- • - 1.404 Rural 0.2789 3.115 5.415 x 10 -• -1.399 (NH4)2504 0.4809 3.082 3.110 X 10 -• -1.428
Reproduced from Table 1 of Gerber [1985]. The temperature de- pendence of (3) is found primarily in C 3 due to the sensitivity of the Kelvin effect to temperature; C3 can be temperature corrected with the
expression: C•(T) : C311 + 0.004(298 - T)], T in kelvin.
Toba and Chean, 1973] and wave tank measurements of the
sea-salt aerosol production rate per unit area of whitecaps,
Monahah et al. [1986] have substantially refined the formula-
tion for the sea-salt aerosol generation and included both in-
direct (bubbles) and direct (spume) mechanisms in the model.
The density function dF/dr (particles m -2 s --• /•m-•), which
expresses the rate of sea-salt droplet generation per unit area of sea surface, per increment of droplet radius, is given (1) for
indirect mechanism (through bubbles):
dFo
dy
--= 1.373U264•r-3(1 + 0.057r l'øs) X 10 (Sa)
where B = (0.380 - log r)/0.650 and (2) for direct mech- anism (through spume)'
0
dF•
8.60 x 10-6e2'øSUlør
-2
dr 4.83 x 10-2e2'øsU•"r -4 8.60 x 106e2'øsU•ør -8 r < 10/zm 10/zm_< r -< 75 75/zm-<r-<100 r _> 100 (5b)The total generation rate of sea-salt aerosols via whitecaps and breaking waves is then
dF dF• dFo
dr = d--•-
+ dr
(6)
A total flux of sea-salt aerosols for each size bin is obtained by integrating this formula over the size range of this bin. This
sea-salt generation model is of essential importance in setting
up the lower boundary conditions for our sea-salt aerosol model. It is noted that the applicable range of radius for (5a) is 0.8-10/zm (at 80% relative humidity (RH)) within which the
empirical relationships are based. It may introduce some un-
certainties if it is used for the generation of submicron sea-salt
aerosols. For particles larger than 10/zm, (5b) should be used
3808 GONG ET AL.: ATMOSPHERIC SEA-SALT MODELING, 1
at high wind speeds compared to observations and is therefore
neglected in the simulation.
Another mechanism of marine aerosol generation, which
has not received much attention during the last 3 decades, is
the production of sea-salt aerosols by wet precipitation. This
mechanism is mentioned here for two reasons. First, since it is
well recognized that a major self-cleaning process of the atmo- sphere is the removal of particles by clouds and precipitation, any attempt to model aerosol particles in the marine atmo-
sphere without incorporating wet removal mechanism is not
realistic. Second, in the case of wet precipitation, if only the removal process is considered, which is the case for all the
present marine aerosol models, aerosol concentrations may be
underestimated, especially near the sea surface.
Unfortunately, literature on this subject is very limited. Only
two papers have been found that deal exclusively with it [Blan-
chard and Woodcock, 1957; Marks, 1990]. In the first paper,
Blanchard and Woodcock [1957] suggest two mechanisms of
marine aerosol generation by raindrops: (1) direct production
through the impact of the rain drops and (2) from bubbles produced in the surface water as a result of impact. This pro-
cess occurs when a rain drop hits the sea surface, it generates
not only sea-salt aerosols and bubbles but also secondary drops
due to the splash. The secondary drops fall back into the sea
and generate more bubbles. It is believed that in both cases the
production of bubbles is a function of droplet size [Blanchard
and Woodcock, 1957]. The generation of sea-salt aerosol par-
ticles follow the bursting of bubbles.
Marks [1990] carried out an experimental investigation of
the marine aerosols under both dry and wet (rain) conditions
from the Dutch research platform Noordwijk, 9 km offshore in
the North Sea. Total sea-salt concentrations were measured at
three heights of 4.5, 12, and 18.3 m and sea-salt size distribu-
tions at 12-m elevation. Compared to dry conditions, a reduc-
tion in the aerosol concentration at 4.5 m by rain was found at
wind speeds of 8 and 10 m s-•. In other words, the net effect
of rain on the total aerosol concentrations at these wind speeds
is negative, an indication that the wet removal process sur-
passes the generation of sea-salt aerosols by rain. When the
wind speed at 4.5 m reached 15 m s -•, the sea-salt aerosol
concentration during rainy weather became greater than under
dry conditions. This means rain generates more sea-salt aero-
sols than it scavenges provided that wind-mediated sea-salt
production remains unchanged during dry or wet conditions. One interesting point from the experiment was that for 12- and
18.3-m elevations at a wind speed of approximately 15 m s -•,
the sea-salt aerosol concentrations were not influenced by rain, meaning that a balance between production and washout of sea-salt aerosols by rain occurred under this condition. If this
is a true characteristic of the marine boundary layer sea-salt
aerosols, it might be used to parameterize processes of rain- mediated generation of sea-salt aerosol. In this respect, wet
scavenging mechanism which have been studied extensively
[Wang et al., 1978; Herbert and Beheng, 1986] would play an important role.
Since the rain intensity was not reported in the paper, a quantitative relationship between the rain-mediated produc-
tion and scavenging and wind speed can not be inferred from
the paper. Therefore this mechanism of sea-salt generation is
neglected in our model. One justification for this would be that in comparison to wind-generated sea-salt aerosols, this source is likely a minor one, especially considering the frequency of the precipitation.
2.3. Wet Removal of the Marine Aerosols
For modeling atmospheric aerosols, the inclusion of wet removal process, that is, the scavenging of particles in the air by precipitation and by clouds, is essential. This self-cleaning
mechanism maintains the balance between the sources and
sinks of atmospheric aerosol particles. In the present model,
both below-cloud and in-cloud scavenging processes are con-
sidered with a simplified approach suitable for use in a large- scale transport model.
2.3.1. Below-cloud scavenging. Below-cloud scavenging
as referred to here is the process of aerosol removal from the
atmosphere between cloud-base and the ground by precipita-
tion (i.e., rain or snow). The capture of aerosol particles by
falling hydrometeors takes place by BrownJan and turbulent
shear diffusion, inertial impaction, diffusiophoresis, thermo-
phoresis, and electrical effects. Detailed studies of these pro- cesses [GreenfieM, 1957; Pilat, 1975; Wang et al., 1978; Slinn,
1984; Herbert and Beheng, 1986] have revealed that there exists
a minimum in the collection efficiency of aerosol particles
between sizes ranging from 0.5 to 1 txm radius, which is some- times called the "Greenfield gap." According to Slinn [1984],
the removal rate of aerosols per unit volume can be written as
L (7)
where r i is the averaged actual radius of ith size bin and ½ is the
scavenging rate which is computed by integration of the fol-
lowing equation over all hydrometeor sizes:
½(r•) = E(r•, a)Ax[Vt(a) - Vg(r•)]N(a) da
where a is the drop radius, Vt and Vg are the gravitational
settling velocity of hydrometeors and particles, E (ri, a) is the collection efficiency of aerosols of size r i by hydrometeors of
size a, and N(a) and Ax are the hydrometeor number density
and effective cross-sectional area. However, the requirement of N(a) and large computer time for the integration makes it unrealistic to use in a large-scale aerosol model. An approxi-
mate expression [Slinn, 1984] for the rain scavenging rate is
used:
cpE(r•, gin)
½(lrt)
=
gr
n
(8)
where c is a numerical factor of order unity (0.5) and the mean
collision efficiency E(ri, Rm) is evaluated by using a mean
drop size. The precipitation rate p is also used to obtain the mean drop radius'
Rm
--'
0.3
5 mm 1 mm/h
(9)
For snow scavenging,
Tp• (r,, A)
qt(rt)
= Dm
(10)
where X is the characteristic capture length, D m is the charac-
teristic length, and T is a constant of order unity (0.6). De-
pending on temperature, X and D m may have different values [Slinn, 1984] because of different type of snows (Table 3).
2.3.2. In-cloud processes. In-cloud scavenging processes
are reviewed by Bartie [1991]. Scavenging of chemical pollut-
GONG lET AL.' ATMOSPHERIC SEA-SALT MODELING, 1 $809
paper. A three-process model was developed to describe the
in-cloud processes of the aerosol particles. The first process is
the activation of sea-salt aerosols, that is, served as cloud
condensation nuclei (CCN), depending on the critical super-
saturation and particle dry size. The mechanism of this process
has been studied in detail by cloud physicists [e.g., Twomey,
1977] and is represented by the so-called K6hler curves. For
sea salt (NaC1) and sulphate ((NH4)2504) , the relationship between the critical radius (rc) of dry soluble particle and
supersaturation (So in percent) is given approximately by
[Twomey, 1977]
Sea salt
rc = 1.16 x 10 -6 S• -2/3
Table 3. Parameters for Snow Scavenging
Crystal Type Din, cm
Graupel particles 1.4 X 10 -2
Rimed plates and stellar dendrites 2.7 x 10 -3 Powder snow and spatial dendrites 1.0 x 10 -3
Plane dendrites 3.8 X 10 -4
Needles 3.8 x 10 -3
Crystal Type X,/am a
sleet, graupel 1000 rimed crystals 100 powder snow 50 dendrites 10 tissue paper 50 camera film 1000 2/3 2/3 2/3 1 Sulphate r•. = 1.40 x 10 -6 S• -2/3
which simply describes the minimum dry radius which would
enable nucleation to occur upon it at a supersaturation So. The
activation process has two major impacts: (1) because of the
larger size of CCN they will be subject to a different removal scheme for particles with size greater than the critical radius r c and hence the size distribution spectrum of sea-salt aerosols
will be modified and (2) a link is established between aerosols
and their indirect effect on climate for the activation will alter
the cloud number density, liquid water content, and possibly
cloud albedo.
The second in-cloud process is the attachment of the aerosol particles in size range i to the existing cloud droplets at a rate a,r. The attachment of aerosols to the cloud droplets is as-
sumed to be by two processes: Brownian and turbulent diffu-
sion; that is, a,r = a,• + a•r•. The contribution by Brownian diffusion to a•r is estimated by [Dingle and Lee, 1973]
kT
( l + al/r,
l + al/rc,)
= --+ (r, + r•)M• (11)
OltB •
Fcl
Ft
where r i is the averaged aerosol particle radius at size bin i and
r• is the cloud equivalent droplet radius; a is the Cunningham
correction factor; I is the mean free path of air molecules; M c is the number density of cloud droplets; k is Boltzmann's constant; T is the absolute temperature; and • is the air vis- cosity. The turbulent contribution [Levich, 1962] is approxi-
mated as
a,rB = 14.1 (r, + r•) 3 M• (12)
where • is the rate of energy dissipation and v is the kinematic
viscosity. One approximation in formulating (11) and (12) is
that the cloud size spectra is replaced by an equivalent droplet radius (rcl), which is estimated by the cloud scheme of the
climate model.
The final process involves the removal of the aerosol-
containing cloud droplets produced by the first two processes
by large falling hydrometeors at a rate /3i [Dingle and Lee,
1973; Slinn, 1984]. Assuming that an activation rate is Ai for size range i and that a,r and /3 i are constant within the inte-
gration interval, the in-cloud processes can be expressed by
following ordinary differential equations:
d Xia
dt :--(OttT q- Ai)Xta -- •/iaXia
d Xic
dt
--= (o•,r + A,)X,.- 13,x,•
(•3)
where mia is the interstitial concentration of the aerosol parti- cles of i th size bin, Xic is the concentration of the aerosols in cloud water, and •/ia is the rate of direct aerosol removal by
large hydrometeors which can usually be ignored when com-
pared to OliT and [3 i. The activation rate A i is determined by the critical radius of aerosol particles. The total aerosol con- centration, Xit = Xia q- Xic, is obtained by solving (13) for
subject to xitl•=o = ?rio.
X,,
=/3,- (c•,r
+ A,)
[/3'e-("•+^')'
- (0lit
q-
A')e-•"]
In the case of nonactivation, that is, r i < r c, A i = O, (14a)
mto
--• [ [3 i e - •" rt -- ot t re - t3't ] (14 b )
Xit-- [3
i __
Otir
and for the activation case, that is, ri -> rc, if we assume A i -- 0%
,)('it -- Xioe-t3,t (14C)
In deriving (14c), an implicit assumption is made that once an
aerosol particle is activated, all the particles with the same dry radius will form cloud droplets (Ai = o•), and they will be removed from the atmosphere if there is precipitation. The removal rate by large hydrometeors,/37, is calculated by setting /3i = ½(rc/) in (8). If no precipitation is experienced during the
time step, the aerosol particle remains in the atmosphere and
total concentration does not change.
2.3.3. Collection etficiency (E). In order to calculate the
scavenging rate of aerosols or clouds by both rain and snow
(equations (8) and (10)), the collection efficiency of aerosol
particles by hydrometeors (E) is required. The collection ef-
ficiency is defined as the ratio of the total number of collisions occurring between droplets and particles to the total number of particles in an area equal to the droplet's effective cross- sectional area. According to previous investigations [Slinn and Hales, 1971; Pilat, 1975; Wang et al., 1978; Herbert and Beheng,
1986], E is a result of the combined action of Brownian diffu- sion, inertial impaction, interception, thermophoresis and dif- fusiophoresis, and electric forces. Detailed formulation of
3810 GONG ET AL.: ATMOSPHERIC SEA-SALT MODELING, 1
•h
Mean
Wind
Speed,
u
SURFACE
LAYER Sea-salt
concentration,
Zi
/
KsL
• turbulent
transfer
velocity ß ß
ß
Ka,
drift
velocity
Friction
velocity,
u,
h
Bounce-off factor, f•o I
/
INTERFACIAL LAYER Zo- surface roughness
ture of the surface roughness are considered only in the inter- facial layer which is from surface to height 8. Assuming steady state, the dry deposition flux of aerosol particles through the
Earth-atmosphere is expressed by Giorgi [1986]:
4i) h --- l•' depFl h (17)
where
nh is the particle
concentration
at height
h and l•'dep,
the
particle dry deposition velocity, is given by Giorgi [1986] as'
•'dep = Kdep + gdh
where the definition of Kdh is given in Figure 2 with
Ksœ(KizfBo + Kda- gdh)
Kdep
= KsL
+ KizfBo
+ Kd•
(18)
K,L - interfacial layer transfer velocity K• - drift velocity
Figure 2. Schematic illustration of two-layer model used for
dry deposition of aerosol particles on water surface. It is as-
sumed a zero surface concentration and a bounce-off factor
fBO.
tively difficult undertaking. Neglecting phoresis and electric
effects, Slinn [1984] derived the following semiempirical ex-
pression for the particle/drop collection efficiency:
4
E = ReSc
[1 + 0.4Re•/2Sc
m + 0.16Re•/2Scm]
[ St-S*
] 3/2
+ 4•f[to
-• + (1 + 2Re•/2)•f]
+ St- S, + 2/3
(15)
where Re = R m Vt/p is the Reynolds number of hydrometeors,
Sc = •,/D and St = [r(Vt - Vg)]/R
m are the Schmidt
and
Stokes numbers of particles. The coefficient r is the character- istic relaxation time of the particles, D is the diffusion coeffi-
cient of particles, Vt and Va are the terminal velocity of hy-
drometeors and particles, and R m and r• are the radius of
hydrometeors and particles, respectively. •15 = ri/Rm, to =
IXw/•',•, and/Xw and/x, are the viscosity of water and hydrom-
eteors.
12/10 + (1/12)Ln(1 + Re) S, =
1 +Ln(1 +Re)
For snow scavenging, the efficiency is estimated as
E= •-• + 1-exp[-(1
+Re}/2)]•j
St- S,
+ St- S, + 2/3
(16)
Depending on the type of snow crystals, a and X values from
Table 3 are used in (16).
2.4. Dry Deposition
In modeling particle dry deposition, a two-layer hypothesis is usually assumed [Slinn and Slinn, 1981; Giorgi, 1986]. Figure 2 shows a typical two-layer model for the water surface. The transport of particles in the top layer, that is, the surface layer extending between h and 8, is dominated by turbulence and gravitational settling, respectively. Viscous effects and the na-
where KsL and KiL are the transfer velocity in surface and
interfacial layer, respectively, and are assumed to be of the
following forms [Giorgi, 1986]'
Ksz•
= CoUhAp
r.•/2_
•.•DOC•2
(19)
Ki/_, = C•02u, G
where Co and Coo are the drag coefficients relative to h and
8, Uh is the wind speed at h. C r• is predicted in the climate model as the drag coefficient for the first layer above the surface and Coo is calculated in the dry deposition module according to the surface type, that is, Zo. The model distin- guishes between the following surfaces: water (ocean or lake), sea-ice, snow and vegetative canopies, and computes the roughness length Zo for them respectively. The friction velocity u, is calculated as
lg , '-- C•2/Uh
G is a function accounting for Brownian diffusion, intercep-
tion, and impaction of particles on the surface elements. Given the surface type, wind speed, air density, temperature at the
ground level, and the surface momentum drag coefficient, the
dry deposition velocity of aerosol particles can be estimated.
Slinn and Slinn [1981] derived a similar expression for dry
deposition of particles on a natural water surface. Hygroscopic growth of particles was taken into account in the saturated viscous sublayer near the surface. As a first approximation, this effect was neglected because of high relative humidity (>80%) in open ocean marine boundary layer.
2.5. Vertical Turbulent Transport
Vertical transport of aerosol particles in the free tropo- sphere due to turbulence is realized using eddy diffusivity for-
mulations in the free atmosphere and drag coefficient formu-
lations at the surface. This method is used in the Canadian
climate model for the transport of momentum, heat, and mois-
ture [McFarlane et al., 1992]. The eddy diffusivities for momen-
tum and heat are of the form
(gm, gh) = 1210V/Ozl(fm, fh) (20)
where V is the horizontal velocity vector and z the distance
above the local terrain. The Richardson number (Ri) depen-
GONG ET AL.: ATMOSPHERIC SEA-SALT MODELING, 1 3811
(fm,
• - •0 Ril/(• + •0 Ri/87 1/2)
Ri < 0
= (1-5eRi)2/(1 + 10(1- e)Ri) O<-Ri<- 1/5• (21)
0 Ri > 1/5•
where
I = kz/(1 + kz/X)
where k is the von K/trm/tn constant and the asymptotic mixing
length
(X) is specified
as 100 m. Like moisture
in the climate
model, the aerosol particles have a diffusivity equal to that for
heat. The flux in the free atmosphere A p-r is written as
- - (22)
Avr
pKa
Oz
where K a -- K e = Kh; Xi and 9 are defined as in (1). The flux
at and near the surface (A •r)x is expressed as
(Avr)s: -9• VL (XL+i/2- xg)Ca (23)
where
Ca = Ch; ,•z•
+ •/: is the lowest
prognostic
level aerosol
mass mixing ratio. The drag coefficient (C•) is related to the
surface-layer
bulk Richardson
number (RIB) by following
equations: Ct• = CDHFt• where F•(RiB)
1 - 10 RiB/(1 + 10 RiB/(87A•)
1/:),
: (1 - 5eRiB)2/[1 + 10(1 -- e)RiB], 0, RIB<0 0 --< RiB --< 1/5e RiB > 1/5e (24)A•l = [(ZO/ZL)1/2]k2/CD.
, Z o is the roughness
height
which
is
calculated inside the climate model, and e is a constant ac- counting for the effects of lack of homogeneity. In the freeatmosphere
e - 3, and at the surface
over water while over
land and ice covered surface, it is set to zero. Transport of
aerosol
particles
by vertical
advection
and convective
mixing
is
not included in this column version simulation.
3. Sea-Salt Aerosol Simulation by the Model
In order to test the aerosol model parameterization, a single
one-dimensional, time variant column version (FIZ-C) [Ther-
rien, 1993]
of the Canadian
general
climate
model GCMII is
chosen
to generate
a climate
with necessary
meteorological
parameters
for the aerosol
model.
This local
climate
model
is
driven by dynamic tendencies (OU/Ot, OV/Ot, Oq/Ot, OT/Ot, 0
In (ps)/Ot) precalculated
by the GCMII at its lateral
boundary
conditions, where U and V denote the horizontal velocities,
while
q, T, and
ps are the specific
humidity,
temperature,
and
surface
pressure,
respectively.
In this
version,
dynamic
tenden-
cies are updated
every
24 hours
during
the simulation.
The
complete
GCM physics
is recalculated
in FIZ-C with 20-min
time step to produce time variant fields such as wind, temper- ature, relative humidity, cloud coverage, and precipitation. The FIZ-C has 10 vertical levels stretching from surface to -34 km
and represent
one grid cell (3.75
ø x 3.75
ø, that is, -400 x 400
km) of the GCM. Since
the dynamic
tendencies
for the tracers
(aerosols)
are not yet available
from the GCM, the horizontal
•
/10
øW
Greenla•/•
/
ocati/ lo øNr
/
Figure 3. The geographic location for the test run in the
North Atlantic.
advection of aerosols is neglected. In other words, it is assumed that there is an infinite homogeneous domain in the horizontal
and the solution for aerosol concentration represents a quasi-
steady state balance between local production and local re-
moval.
A grid
point
from the GCMII in the North Atlantic
(latitude
61.2 ø north, longitude 352.5 ø ) was selected for the test run
(Figure
3). The simulation
was
run for 90 days
starting
January
1. The aim was to evaluate the performance of the model by
comparing
model
predictions
with available
experimental
ob-
servations from that region. Because of the one-dimensional nature of the FIZ-C used with the aerosol model, the compar-
ison shall not serve as a validation analysis which will be per- formed when the aerosol model is coupled to a three-
dimensional regional or global version of GCMII but rather it will serve to illustrate qualitatively how realistic are the param- eterizations of various processes affecting sea-salt aerosols.
Figure 4 shows
simulation
results
for aerosols
in the four
lowest atmospheric layers (layer depth: 0-166 m, 167-722 m, 723-1767 m, and 1767-3594 m) together with the wind speed at 10 m and precipitation for January, February, and March. Predictions for the first 10 days were omitted. This is the
spin-up
time for the model.
Note that the simulation
region
in
the North Atlantic is on the storm track of cyclones during
winter. The periodic variation of wind and precipitation shows a succession of synoptic disturbances with a period of 3-5 days characteristic of winter storms. The variation of all meteoro- logical variables is consistent with those in the GCM. Compar- ison of the two graphs immediately reveals that the sea-salt concentration is regulated more strongly by surface wind than
by precipitation,
especially
the surface
layer (0-166 m). De-
tailed discussion of the results will be given in the following
sections.
3.1. Observation Available for Model Validation
A summary of observational data for sea-salt aerosols is listed in Table 4 for comparison with the model prediction.
Experimental
location,
measurement
method,
sampling
height
as well as the maximum wind speed during the measurement
are given
in the table.
There are basically
two parameters
for
determining the sea-salt aerosols, that is, their size and mass.
3812 GONG ET AL.' ATMOSPHERIC SEA-SALT MODELING, 1
0-166m ... 166 - 722 rn
a) Four Layer Sea-salt --- 722 - 1767 rn
--- 1767 - 3594 rn 103 ...
,•'
10'
.', ' I•'
:': :' \ll ,'
• 10
ø ' ",.'
", • "" "';'
:: ø'-:
':" "::
':" :"
': • '•-:
: ' '" I ,
10"
j / '"'
ß .' lO-4 10 -5 .... 10 20 30 40 50 60 70 80 90 Julian Dayb) Wind Speed & Precipitation
25 0.0006 • 20 '• 15 • 10 ._ E o 5 0 • 0.0005 •E 0.0004 .• 0.0003 •. 0.0002 o. 0.0001 0.0000 10 20 30 40 50 60 70 80 90 Julian Day
Figure 4. Simulation results for the test location. (a) Varia-
tion of total sea-salt aerosol concentrations in the first four
model layers extending from the surface to 3594 m. (b) The
wind speed and precipitation for the same period.
and Marks [1990] measured the total sea-salt mass as a func-
tion of wind speed. Their methods involve the collection of marine particles on a filter and then analyze the sodium con- tent on it. While Marks obtained the total sodium by neutron activation method, the others measured only the soluble so-
dium. The total Na measurement eliminates the artifacts
caused by non-sea-salt particles. Particle size distribution was
measured by Woodcock [1953], Extort et al. [1986], Gras and
Ayers [1983], and Marks [1990] by different techniques. Sea-salt
content was obtained by analyzing the sodium [Woodcock, 1953] and by using the sea-salt density and particle volume [Extort et al., 1986; Gras and Ayers, 1983]. The latter technique for obtaining sea-salt contents may be subject to interference
by non sea salt in the particles. An important assumption in
this particle size to sea salt mass calculation is that the col- lected aerosol sample is close to 100% marine origin.
The location of experiment can strongly affect the represen-
tativeness of observations of open marine aerosols modeled
here. Measurements conducted on a ship over the ocean [Tsu-
nogai et al., 1972; Lovett, 1978; Marks, 1990] are closer to ideal
marine aerosols than those performed inland [Kulkami et al.,
1982] or on a coast [Gras and Ayers, 1983; Extort et al., 1986].
The inland (1.8 km) observation by Kulkarni et al. was cer-
tainly affected by coastal surf as well as by dry depositional
losses and hence is not appropriate for comparison with our
model. Even if the measurement was made over open ocean,
the particle size is not completely characteristic of sea-salt
aerosols. Anthropogenic sulphates and formation of DMS ox-
idation products can contribute substantial into non-sea-salt
fine particles, especially during high biogenic activity periods. Therefore the total sea-salt mass from particle size measure-
ments [Extort et al., 1986; Gras and Ayers, 1983] may be sub-
jected to some uncertainties. It also seems that the measure- ments by Gras and Ayers and Exton et al. may suffer from beach surf effects, sedimentation, and turbulent deposition losses to the beach as their probes were several hundred meters away from the tide line.
Most of the measurements in Table 4 were made below
100 m, except Woodcock [1953] whose data were taken at cloud
levels (-600 m). Accordingly, the observations by Tsunogai et al. [1972], Lovett [1978], and Marks [1990] are chosen for com- parison with our model predictions for the surface layer (0-166
m), while that by Woodcock [1953] is used for comparison to
simulation in layer 2 (167-722 m). The total sea-salt by sodium
measurements and marine test locations by these investigators
match closely to the conditions used in the simulation.
3.2. Size Distribution of Sea-Salt Aerosols
The particle radius ranging between 0.03 /am and 8/am is divided into 8 size bins according to Table 5. In this size range,
the main mechanism of sea-salt generation is attributed to the
Table 4. Published Constants for the Relationships Between Airborne Sea-Salt Concentration X and Wind Speed U•o by
Using Equation (25)
Maximum
Method of Location and Sampling Wind, a, b,
Authors Latitude Measurements Height m s -• s m -• /•g m -3
Woodcock [1953] 20øN glass slide (Na and size) cloud base over Pacific Ocean 35 0.16 2.57 (600 m)
Tsunogai et al. [1972] 40øN-50øN membrane filter (na) ship on Pacific Ocean 18 0.62 0.33 (12-14 m)
Lovett [1978] 50øN-60øN membrane filter (na) weather ships in Atlantic Ocean 20 0.16 4.26 (5-15 m)
Kulkarni et al. [1982] 19øN membrane filter (na) W. Indian coast 1.8 km inland 8 0.27 5.35 Exton et al. [1986] 57øN ASAS and CSAS (size) Hebridean beach facing Atlantic 20 0.17 14.30
(15 m)
Gras and Ayers [1983] 41øS round jet collector (size) Cape Grim (94 m) 20 0.124 2.52 Marks [1990] N/A neutron activation (impactor Na) ship on North Sea (12 m) 24 0.23 1.10 Marks [1990] N/A neutron activation (filter Na) ship on North Sea (12 m) 24 0.23 1.13
GONG ET AL.: ATMOSPHERIC SEA-SALT MODELING, 1 3813
Table 5. Aerosol Size Discretization Scheme
Bin Size Range
1 0.03-0.06 2 0.06-0.13 3 0.13-0.25 4 0.25-0.50 5 0.50-1 6 1-2 7 2-4 8 4-8
The radius is in micrometers.
indirect scheme, that is, via bubbles. The predicted mass size
distribution of sea-salt aerosol (Figure 5a) is compared with
experimental results of Marks [1990] in the North Sea and McGovern et al. [1994] at Mace Head on the west coast of
Ireland. Several features are revealed by the comparison. The
size distribution (1) of the sea-salt mass concentration com-
pares well within the observational data (2) in terms of the
shapes of the distribution curves and the magnitudes of the
mass concentrations (Figure 5c). This agreement indicates that
the parameterization schemes adapted in the model are per-
formed reasonably well. It is also seen from Figure 5a that
a) PREDICTIONS b) MARKS OBSERVATIONS
10 2 • 101 E • 100 10-1 i 11ms 0 1 2 3 4 5 Dry Radius [gm] 10 ø 10-1 -i .... i .... i .... t .... i .... i- •... ..
/ r
22
rn
s
-1
/ 8ms '1 I .... I .... I .... I .... I .... I 0 1 2 3 4 5Median Radius [Izm]
c) PREDICTION vs McGOVERN et al 10 2 101 • 10 0 10-1
•.._..______-••• Wind
Speed
I Prediction 10-11
ms'
//
Prediction
[--•--
---.--
7ms
170r•1:1
'1 s'
•
• Obse•ation{•
' •lm•-•
i
I = 14ms
'1
10-2 , , , , , ,•,l
, •[ , • • ,,[t
• • , , • •
0.01 0.1 1 10 Radius [l•m]Figure 5. The sea-salt aerosol size distribution of mass concentrations. (a) Model predictions at three wind
speeds; (b) experimental observations [Marks, 1990] for both dry and wet conditions; (c) comparison with
McGovern et al. [1994]. The vertical dashed line in Figure 5c indicates the smallest experimental radius at
3814 GONG ET AL.' ATMOSPHERIC SEA-SALT MODELING, 1
a) PREDICTION b) SURFACE FLUX
109 _--- 109 108 • 108 • 107 • 0 7 • 106 • 106
• 10
5
19.ms'•q '• 10
.•-5
Q • • 103 • 10 •z 10•
10 • 10 • 100 0.01 0.1 1 10 O0 Dry Radius [!•n], r d 10 o 0.01 19 ms -•I• 10.8
8.9 m s '•m
s
'•
4.7 m s '• I I ... i , ,I ... I ... I 0.1 1 10Dry Radius [lu.m], r d
c) Predictions vs Observations 101o 10 9 E .--. 10 8 o :• 10 7 d]ogr, -• .m 10 6 • 10 5 E lO 4 z 10 3 102 101 •-'•_ .-•. • I Prediction '-•" "',"•/ *• I at 19 ms"
'
• '._.
•-•
ß
Prediction
atlOms"• I
I ! I I ! ! ! '. Observations 10 o 0.01 0.1 1 10 100 Radius [[tm] rdFigure 6. (a) As a function of wind speed, the number size distribution of sea-salt aerosol particles predicted by the model. (b) The predicted sea-salt surface flux distribution (equation (5)) corresponding to results in
Figure 6a. (c) Comparison of observations [from Fitzgerald, 1991, Figure 1] with predictions at two wind
speeds. The vertical dashed line in Figures 6b and 6c indicates the lowest experimental radius at which (5) was
based.
some detailed structures of the distribution at submicron
ranges are resolved by the simulation. There is a hint of bi-
modal mass size distribution in the model results. This feature
is even more pronounced when the predictions are plotted as
number size distribution (Figure 6). The bimodal distribution
from the model is not as clear as for the observations. This is
likely due to lower size resolution in the model than observa-
tions as well as the fact that processes of cycling the aerosol
through cloud formation and evaporation are not considered
in the present one-dimensional model. They will, however, be
taken into account in the three-dimensional model to be run
later.
Furthermore, coagulation of aerosols is ignored in this ver- sion of the model. As a result, at fine particle size ranging below 0.1 /•m, the prediction yielded a higher concentration
than the observation for both mass and number size distribu-
tion. Coagulation would also strengthen the concentration near 0.1 /•m and thereby enhance the bimodal structure. An-
other effect is that other non-sea-salt aerosols also contribute
to the distribution measured in the remote marine environ-
ment. However, in the radius range where sea-salt dominates
(0.5 to 10/•m) the model and experimental results agree rea-
sonably well. A comparison of Figures 6a and 6b illustrate that the shape of the number size distribution curves is strongly influenced by the sea surface flux size distribution of sea-salt droplets which is the only source of sea-salt aerosols. As seen from Figure 6b, the flux curve also does not show a distinct
bimodal distribution.
It should be pointed out that the empirical sea-salt produc- tion function by Monahah et al. [1986] (equation (Sa)) is based on observations of sea-salt generation for particles larger than about 0.3/•m in radius (dry). This is the lower limit of particles
GONG ET AL.: ATMOSPHERIC SEA-SALT MODELING, 1 3815
detected in the experiments with which the production func-
tion was compared well. Extrapolation beyond the range re-
sults in uncertainties since the empirical function begins to
deviate from the experimental observations below 0.8 tzm ra-
dius (80% RH). From this model simulation, use of Monahah
et al.'s [1986] production function seems to reasonably simulate
production down to a radius of about 0.1 tzm (dry). However,
below 0.1 tzm, a large overestimate of both mass and number
concentrations occurs (e.g., the prediction curves on the left of a dashed line in Figures 5c and 6c).
3.3. Wind Speed Dependence of Sea-Salt Aerosols
The effect of wind speed on the sea-salt aerosol ambient concentration (X) has been investigated experimentally by many researchers [Woodcock, 1953; Lepple et al., 1983; Lovett, 1978; Gras and Ayers, 1983; Extort et al., 1986; Marks, 1990]. Generally, the dependence is expressed as a log-linear fit to the
experimental data between concentration and wind speed:
In X' = In (b) + auto (25)
In [•'icr•rp '7h rntaclpl nrpctir, t•r•nc c•f textual rn•acc •arp r,c•rnn•arpct
experimental observations in the surface layer. A general
agreement for the wind effect is obtained between the predic-
tions and experiments (Figure 7b). Best fitting curves from the model results are also shown in Figure 7a. A slope (a) 05 0.26
and 0.13 and a correlation coe•cient of 0.82 and 0.23 are
predicted for total and fine particle (r, - 0.13 to 0.25 /•m) mass, respectively. Using a five-stage impactor and filter sam-
pler to measure the sea-salt aerosols at various wind speeds,
Marlcs [1990] obtained the dependence 05 sea-salt aerosols on
wind speed for five •ize ranges and œor total sea salt, respec-
tively. Under dry nonprecipitating conditions, except for the
dinmeter of 0.•9-0.95 •m where the correlation coe•cient is
0.31, the slopes for all other four size stages ranging from 0.95 to 30 /•m (diameter) was about 0.23 [Marlcs, 1990, Table 2], which is very close to the model simulated results of 0.26 for total mass. The slope is 0.12 for the observation of particles in diameter of 0.49-0.95/•m and 0.13 for the simulation of par- ticles in the radii ranging from 0.13 to 0.25/•m.
Exton et al. [1985, Table 3] found the same trends for three size categories of sea-salt aerosols based on the field measure-
ments off the northwest coast of Scotland. For the ultra large
sea-salt (r = 8-16/•m) aerosols, the slope and the correlation
coefficient were 0.235 (average) and 0.59 (average) respec-
tively. For the large (r = 0.3-8/•m) aerosols, they changed to
0.13 and 0.56 respectively. For the small (r = 0.1-0.3 /•m)
aerosols, the slope was reduced even more to 0.08 while the
correlation coefficient was 0.4.
Tsunogai et al. [1972] measured sea-salt concentration from a cruise ship on the Pacific Ocean and a slope of 0.62 was
obtained for the wind dependence. It is much larger than other
observation's. They attributed it to two reasons: (1) the pro-
duction of sea-salt particles > 10/•m in radius or 10- s g in mass
is more dependent on the wind speed than the production of smaller ones; (2) the residence time of the larger particles is so
short that they contribute only to the weight concentration of
sea-salt particles near the surface of the ocean. Lovett [1978] attributed the large value to the moving ship which may gen- erate sea-salt particles. This effect is unlikely true since sam- ples were discarded when a tail wind was recorded. In our opinion, it may be due to the experimental method used to construct the sea-salt wind dependence. It was reported in the
paper that the sampling time was from 20 to 50 hours. During the long experimental period, the wind speed might undergo
substantial change. Because of the exponential relationship
between sea salt and wind speed, the averaged sea-salt con-
centration at the averaged wind speed could be higher than an
averaged sea-salt at the same averaged wind for a short time
when
the wind
speed
would
not change
substantially.
This
is
clearly shown in Figure 8 where the wind speed is assumed to
undergo a sinusoidal change from 0 ø to 180 ø with a speed of
0-15-0 m s-1 cycle for a designated time and an exponential
relationship is assumed between sea salt and wind speed. For
sampling through 0ø-180 ø with every degree a sample (dura-
tion 180), the root mean square (rms) wind speed for the duration is 10.6 m s -•, and the rms sea-salt concentration is
29.6 /•g m -3. For a shorter
sampling
duration
(51) which
re-
sults in the same rms wind speed (10.6 m s-•), the rms sea-salt concentration is reduced to 22.71 tzg m -3. When the wind speed is constant at 10.6 m s- • for the duration (180 or 51), the sea-salt concentration is 16.37 m s -• which is given by the
assumed exponential relationship. Furthermore, even if the
wind speed only changes slightly in a 15-16-15 m s-• cycle, the
long .... •i•,,, ,4 .... ,i,•, (!8{)) still ... •t• in ... tlrn•toct rms sea-salt concentration when compared with a sampling
duration of 51 for the same rms wind speed. Therefore, if wind
speed changes, the long-time average of sea-salt and wind speed by Tsurlogai et al. [1972] actually overestimates the real relationship between the sea salt and wind speed, that is,
higher slope, compared to the sampling time of 0.5 to 3 hours
of Marks [1990] and 1-2 to several hours of Lovett [1978]. The sea-salt and wind speed relationship given by Woodcock [1953] was obtained at cloud levels (about 600 m) in the Pacific with a slope (a) of 0.16. Although Lovett [1978] obtained the same value for an annual average between August 1974 and
July 1975 in the Atlantic, the measurements were performed at
5-15 m. It is expected that as altitude increases the dependency
of sea salt on wind becomes weaker. This feature is reflected in
the model which gives a smaller slope of 0.15 in the layer of
166-722 m than that of 0.26 in the surface layer. It is closer to
0.16 of Woodcock [1953].
The constant b in (25) is referred as the background sea-salt loading. It is the sea-salt concentration when wind speed reaches zero. From Figure 7a, our prediction yields a value of
1.4/•g m -3 for Heimaey.
Compared
to the observations
(Table
4), b is 1.13/•g m --• for Marks
[1990],
2.57/•g m --• for Wood-
cock [1953], 0.33/•g m -'• for Tsunogai et al. [1972], and 4.26/•g m -3 for Lovett [1978]. There are two factors which may explain
the observational difference. (1) The three experiments were
made at the North Atlantic [Lovett, 1978], the North Sea [Marks, 1990] and the Pacific [Tsunogai et al., 1972] respec-
tively. Different meteorology such as wind speed and precipi-
tation at these sites will result in different patterns for sea-salt
generation and removal which influence the constant a and b in (25). Our simulation also shows the variability of constant a
and b at various sites (Table 6). The salinity of the ocean may play a role in the difference but may not account for the big
discrepancy between 4.26 and 0.218 since the difference be-
tween the highest salinity of 35.5%e in the North Atlantic and
the lowest of 34.2%e in the North Pacific cannot contribute
such a big difference in sea-salt concentration; (2) the mea-
surement techniques used, for example, the sampling time,
average methods and equipment. The sampling duration prob-
lem has been discussed above with Tsunogai et al.'s [1972]